From 51e8513615ed8202b22ba9a43b0c7376ea4f6868 Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Thu, 13 Oct 2005 21:10:08 -0700 Subject: [SPARC64]: Consolidate common PCI IOMMU init code. All the PCI controller drivers were doing the same thing setting up the IOMMU software state, put it all in one spot. Signed-off-by: David S. Miller --- include/asm-sparc64/pbm.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include') diff --git a/include/asm-sparc64/pbm.h b/include/asm-sparc64/pbm.h index 38bbbccb4068..c067407de0b0 100644 --- a/include/asm-sparc64/pbm.h +++ b/include/asm-sparc64/pbm.h @@ -102,7 +102,7 @@ struct pci_iommu { u32 dma_addr_mask; }; -extern void pci_iommu_table_init(struct pci_iommu *, int); +extern void pci_iommu_table_init(struct pci_iommu *iommu, int tsbsize, u32 dma_offset, u32 dma_addr_mask); /* This describes a PCI bus module's streaming buffer. */ struct pci_strbuf { -- cgit v1.2.3-71-gd317 From 688cb30bdc3e398d97682a6a58f825821ee838c2 Mon Sep 17 00:00:00 2001 From: "David S. Miller" Date: Thu, 13 Oct 2005 22:15:24 -0700 Subject: [SPARC64]: Eliminate PCI IOMMU dma mapping size limit. The hairy fast allocator in the sparc64 PCI IOMMU code has a hard limit of 256 pages. Certain devices can exceed this when performing very large I/Os. So replace with a more simple allocator, based largely upon the arch/ppc64/kernel/iommu.c code. Signed-off-by: David S. Miller --- arch/sparc64/kernel/pci_iommu.c | 365 +++++++++++++++------------------------- include/asm-sparc64/pbm.h | 28 +-- 2 files changed, 145 insertions(+), 248 deletions(-) (limited to 'include') diff --git a/arch/sparc64/kernel/pci_iommu.c b/arch/sparc64/kernel/pci_iommu.c index cdee7d9fed72..a11910be1013 100644 --- a/arch/sparc64/kernel/pci_iommu.c +++ b/arch/sparc64/kernel/pci_iommu.c @@ -49,12 +49,6 @@ static void __iommu_flushall(struct pci_iommu *iommu) /* Ensure completion of previous PIO writes. */ (void) pci_iommu_read(iommu->write_complete_reg); - - /* Now update everyone's flush point. */ - for (entry = 0; entry < PBM_NCLUSTERS; entry++) { - iommu->alloc_info[entry].flush = - iommu->alloc_info[entry].next; - } } #define IOPTE_CONSISTENT(CTX) \ @@ -80,9 +74,61 @@ static void inline iopte_make_dummy(struct pci_iommu *iommu, iopte_t *iopte) iopte_val(*iopte) = val; } +/* Based largely upon the ppc64 iommu allocator. */ +static long pci_arena_alloc(struct pci_iommu *iommu, unsigned long npages) +{ + struct pci_iommu_arena *arena = &iommu->arena; + unsigned long n, i, start, end, limit; + int pass; + + limit = arena->limit; + start = arena->hint; + pass = 0; + +again: + n = find_next_zero_bit(arena->map, limit, start); + end = n + npages; + if (unlikely(end >= limit)) { + if (likely(pass < 1)) { + limit = start; + start = 0; + __iommu_flushall(iommu); + pass++; + goto again; + } else { + /* Scanned the whole thing, give up. */ + return -1; + } + } + + for (i = n; i < end; i++) { + if (test_bit(i, arena->map)) { + start = i + 1; + goto again; + } + } + + for (i = n; i < end; i++) + __set_bit(i, arena->map); + + arena->hint = end; + + return n; +} + +static void pci_arena_free(struct pci_iommu_arena *arena, unsigned long base, unsigned long npages) +{ + unsigned long i; + + for (i = base; i < (base + npages); i++) + __clear_bit(i, arena->map); +} + void pci_iommu_table_init(struct pci_iommu *iommu, int tsbsize, u32 dma_offset, u32 dma_addr_mask) { - unsigned long i, tsbbase, order; + unsigned long i, tsbbase, order, sz, num_tsb_entries; + + num_tsb_entries = tsbsize / sizeof(iopte_t); /* Setup initial software IOMMU state. */ spin_lock_init(&iommu->lock); @@ -90,27 +136,16 @@ void pci_iommu_table_init(struct pci_iommu *iommu, int tsbsize, u32 dma_offset, iommu->page_table_map_base = dma_offset; iommu->dma_addr_mask = dma_addr_mask; - switch (tsbsize / (8 * 1024)) { - case 64: - iommu->page_table_sz_bits = 16; - break; - case 128: - iommu->page_table_sz_bits = 17; - break; - default: - prom_printf("PCI_IOMMU: Illegal TSB size %d\n", - tsbsize / (8 * 1024)); + /* Allocate and initialize the free area map. */ + sz = num_tsb_entries / 8; + sz = (sz + 7UL) & ~7UL; + iommu->arena.map = kmalloc(sz, GFP_KERNEL); + if (!iommu->arena.map) { + prom_printf("PCI_IOMMU: Error, kmalloc(arena.map) failed.\n"); prom_halt(); - break; - }; - - iommu->lowest_consistent_map = - 1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS); - - for (i = 0; i < PBM_NCLUSTERS; i++) { - iommu->alloc_info[i].flush = 0; - iommu->alloc_info[i].next = 0; } + memset(iommu->arena.map, 0, sz); + iommu->arena.limit = num_tsb_entries; /* Allocate and initialize the dummy page which we * set inactive IO PTEs to point to. @@ -132,114 +167,24 @@ void pci_iommu_table_init(struct pci_iommu *iommu, int tsbsize, u32 dma_offset, } iommu->page_table = (iopte_t *)tsbbase; - for (i = 0; i < tsbsize / sizeof(iopte_t); i++) + for (i = 0; i < num_tsb_entries; i++) iopte_make_dummy(iommu, &iommu->page_table[i]); } -static iopte_t *alloc_streaming_cluster(struct pci_iommu *iommu, unsigned long npages) +static inline iopte_t *alloc_npages(struct pci_iommu *iommu, unsigned long npages) { - iopte_t *iopte, *limit, *first; - unsigned long cnum, ent, flush_point; - - cnum = 0; - while ((1UL << cnum) < npages) - cnum++; - iopte = (iommu->page_table + - (cnum << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS))); - - if (cnum == 0) - limit = (iommu->page_table + - iommu->lowest_consistent_map); - else - limit = (iopte + - (1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS))); + long entry; - iopte += ((ent = iommu->alloc_info[cnum].next) << cnum); - flush_point = iommu->alloc_info[cnum].flush; - - first = iopte; - for (;;) { - if (IOPTE_IS_DUMMY(iommu, iopte)) { - if ((iopte + (1 << cnum)) >= limit) - ent = 0; - else - ent = ent + 1; - iommu->alloc_info[cnum].next = ent; - if (ent == flush_point) - __iommu_flushall(iommu); - break; - } - iopte += (1 << cnum); - ent++; - if (iopte >= limit) { - iopte = (iommu->page_table + - (cnum << - (iommu->page_table_sz_bits - PBM_LOGCLUSTERS))); - ent = 0; - } - if (ent == flush_point) - __iommu_flushall(iommu); - if (iopte == first) - goto bad; - } - - /* I've got your streaming cluster right here buddy boy... */ - return iopte; + entry = pci_arena_alloc(iommu, npages); + if (unlikely(entry < 0)) + return NULL; -bad: - printk(KERN_EMERG "pci_iommu: alloc_streaming_cluster of npages(%ld) failed!\n", - npages); - return NULL; + return iommu->page_table + entry; } -static void free_streaming_cluster(struct pci_iommu *iommu, dma_addr_t base, - unsigned long npages, unsigned long ctx) +static inline void free_npages(struct pci_iommu *iommu, dma_addr_t base, unsigned long npages) { - unsigned long cnum, ent; - - cnum = 0; - while ((1UL << cnum) < npages) - cnum++; - - ent = (base << (32 - IO_PAGE_SHIFT + PBM_LOGCLUSTERS - iommu->page_table_sz_bits)) - >> (32 + PBM_LOGCLUSTERS + cnum - iommu->page_table_sz_bits); - - /* If the global flush might not have caught this entry, - * adjust the flush point such that we will flush before - * ever trying to reuse it. - */ -#define between(X,Y,Z) (((Z) - (Y)) >= ((X) - (Y))) - if (between(ent, iommu->alloc_info[cnum].next, iommu->alloc_info[cnum].flush)) - iommu->alloc_info[cnum].flush = ent; -#undef between -} - -/* We allocate consistent mappings from the end of cluster zero. */ -static iopte_t *alloc_consistent_cluster(struct pci_iommu *iommu, unsigned long npages) -{ - iopte_t *iopte; - - iopte = iommu->page_table + (1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS)); - while (iopte > iommu->page_table) { - iopte--; - if (IOPTE_IS_DUMMY(iommu, iopte)) { - unsigned long tmp = npages; - - while (--tmp) { - iopte--; - if (!IOPTE_IS_DUMMY(iommu, iopte)) - break; - } - if (tmp == 0) { - u32 entry = (iopte - iommu->page_table); - - if (entry < iommu->lowest_consistent_map) - iommu->lowest_consistent_map = entry; - return iopte; - } - } - } - return NULL; + pci_arena_free(&iommu->arena, base >> IO_PAGE_SHIFT, npages); } static int iommu_alloc_ctx(struct pci_iommu *iommu) @@ -279,7 +224,7 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad struct pcidev_cookie *pcp; struct pci_iommu *iommu; iopte_t *iopte; - unsigned long flags, order, first_page, ctx; + unsigned long flags, order, first_page; void *ret; int npages; @@ -297,9 +242,10 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad iommu = pcp->pbm->iommu; spin_lock_irqsave(&iommu->lock, flags); - iopte = alloc_consistent_cluster(iommu, size >> IO_PAGE_SHIFT); - if (iopte == NULL) { - spin_unlock_irqrestore(&iommu->lock, flags); + iopte = alloc_npages(iommu, size >> IO_PAGE_SHIFT); + spin_unlock_irqrestore(&iommu->lock, flags); + + if (unlikely(iopte == NULL)) { free_pages(first_page, order); return NULL; } @@ -308,31 +254,15 @@ void *pci_alloc_consistent(struct pci_dev *pdev, size_t size, dma_addr_t *dma_ad ((iopte - iommu->page_table) << IO_PAGE_SHIFT)); ret = (void *) first_page; npages = size >> IO_PAGE_SHIFT; - ctx = 0; - if (iommu->iommu_ctxflush) - ctx = iommu_alloc_ctx(iommu); first_page = __pa(first_page); while (npages--) { - iopte_val(*iopte) = (IOPTE_CONSISTENT(ctx) | + iopte_val(*iopte) = (IOPTE_CONSISTENT(0UL) | IOPTE_WRITE | (first_page & IOPTE_PAGE)); iopte++; first_page += IO_PAGE_SIZE; } - { - int i; - u32 daddr = *dma_addrp; - - npages = size >> IO_PAGE_SHIFT; - for (i = 0; i < npages; i++) { - pci_iommu_write(iommu->iommu_flush, daddr); - daddr += IO_PAGE_SIZE; - } - } - - spin_unlock_irqrestore(&iommu->lock, flags); - return ret; } @@ -342,7 +272,7 @@ void pci_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_ struct pcidev_cookie *pcp; struct pci_iommu *iommu; iopte_t *iopte; - unsigned long flags, order, npages, i, ctx; + unsigned long flags, order, npages; npages = IO_PAGE_ALIGN(size) >> IO_PAGE_SHIFT; pcp = pdev->sysdata; @@ -352,46 +282,7 @@ void pci_free_consistent(struct pci_dev *pdev, size_t size, void *cpu, dma_addr_ spin_lock_irqsave(&iommu->lock, flags); - if ((iopte - iommu->page_table) == - iommu->lowest_consistent_map) { - iopte_t *walk = iopte + npages; - iopte_t *limit; - - limit = (iommu->page_table + - (1 << (iommu->page_table_sz_bits - PBM_LOGCLUSTERS))); - while (walk < limit) { - if (!IOPTE_IS_DUMMY(iommu, walk)) - break; - walk++; - } - iommu->lowest_consistent_map = - (walk - iommu->page_table); - } - - /* Data for consistent mappings cannot enter the streaming - * buffers, so we only need to update the TSB. We flush - * the IOMMU here as well to prevent conflicts with the - * streaming mapping deferred tlb flush scheme. - */ - - ctx = 0; - if (iommu->iommu_ctxflush) - ctx = (iopte_val(*iopte) & IOPTE_CONTEXT) >> 47UL; - - for (i = 0; i < npages; i++, iopte++) - iopte_make_dummy(iommu, iopte); - - if (iommu->iommu_ctxflush) { - pci_iommu_write(iommu->iommu_ctxflush, ctx); - } else { - for (i = 0; i < npages; i++) { - u32 daddr = dvma + (i << IO_PAGE_SHIFT); - - pci_iommu_write(iommu->iommu_flush, daddr); - } - } - - iommu_free_ctx(iommu, ctx); + free_npages(iommu, dvma, npages); spin_unlock_irqrestore(&iommu->lock, flags); @@ -418,25 +309,27 @@ dma_addr_t pci_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direct iommu = pcp->pbm->iommu; strbuf = &pcp->pbm->stc; - if (direction == PCI_DMA_NONE) - BUG(); + if (unlikely(direction == PCI_DMA_NONE)) + goto bad_no_ctx; oaddr = (unsigned long)ptr; npages = IO_PAGE_ALIGN(oaddr + sz) - (oaddr & IO_PAGE_MASK); npages >>= IO_PAGE_SHIFT; spin_lock_irqsave(&iommu->lock, flags); + base = alloc_npages(iommu, npages); + ctx = 0; + if (iommu->iommu_ctxflush) + ctx = iommu_alloc_ctx(iommu); + spin_unlock_irqrestore(&iommu->lock, flags); - base = alloc_streaming_cluster(iommu, npages); - if (base == NULL) + if (unlikely(!base)) goto bad; + bus_addr = (iommu->page_table_map_base + ((base - iommu->page_table) << IO_PAGE_SHIFT)); ret = bus_addr | (oaddr & ~IO_PAGE_MASK); base_paddr = __pa(oaddr & IO_PAGE_MASK); - ctx = 0; - if (iommu->iommu_ctxflush) - ctx = iommu_alloc_ctx(iommu); if (strbuf->strbuf_enabled) iopte_protection = IOPTE_STREAMING(ctx); else @@ -447,12 +340,13 @@ dma_addr_t pci_map_single(struct pci_dev *pdev, void *ptr, size_t sz, int direct for (i = 0; i < npages; i++, base++, base_paddr += IO_PAGE_SIZE) iopte_val(*base) = iopte_protection | base_paddr; - spin_unlock_irqrestore(&iommu->lock, flags); - return ret; bad: - spin_unlock_irqrestore(&iommu->lock, flags); + iommu_free_ctx(iommu, ctx); +bad_no_ctx: + if (printk_ratelimit()) + WARN_ON(1); return PCI_DMA_ERROR_CODE; } @@ -527,10 +421,13 @@ void pci_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int struct pci_iommu *iommu; struct pci_strbuf *strbuf; iopte_t *base; - unsigned long flags, npages, ctx; + unsigned long flags, npages, ctx, i; - if (direction == PCI_DMA_NONE) - BUG(); + if (unlikely(direction == PCI_DMA_NONE)) { + if (printk_ratelimit()) + WARN_ON(1); + return; + } pcp = pdev->sysdata; iommu = pcp->pbm->iommu; @@ -556,13 +453,14 @@ void pci_unmap_single(struct pci_dev *pdev, dma_addr_t bus_addr, size_t sz, int /* Step 1: Kick data out of streaming buffers if necessary. */ if (strbuf->strbuf_enabled) - pci_strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction); + pci_strbuf_flush(strbuf, iommu, bus_addr, ctx, + npages, direction); - /* Step 2: Clear out first TSB entry. */ - iopte_make_dummy(iommu, base); + /* Step 2: Clear out TSB entries. */ + for (i = 0; i < npages; i++) + iopte_make_dummy(iommu, base + i); - free_streaming_cluster(iommu, bus_addr - iommu->page_table_map_base, - npages, ctx); + free_npages(iommu, bus_addr - iommu->page_table_map_base, npages); iommu_free_ctx(iommu, ctx); @@ -667,6 +565,8 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int pci_map_single(pdev, (page_address(sglist->page) + sglist->offset), sglist->length, direction); + if (unlikely(sglist->dma_address == PCI_DMA_ERROR_CODE)) + return 0; sglist->dma_length = sglist->length; return 1; } @@ -675,21 +575,29 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int iommu = pcp->pbm->iommu; strbuf = &pcp->pbm->stc; - if (direction == PCI_DMA_NONE) - BUG(); + if (unlikely(direction == PCI_DMA_NONE)) + goto bad_no_ctx; /* Step 1: Prepare scatter list. */ npages = prepare_sg(sglist, nelems); - /* Step 2: Allocate a cluster. */ + /* Step 2: Allocate a cluster and context, if necessary. */ spin_lock_irqsave(&iommu->lock, flags); - base = alloc_streaming_cluster(iommu, npages); + base = alloc_npages(iommu, npages); + ctx = 0; + if (iommu->iommu_ctxflush) + ctx = iommu_alloc_ctx(iommu); + + spin_unlock_irqrestore(&iommu->lock, flags); + if (base == NULL) goto bad; - dma_base = iommu->page_table_map_base + ((base - iommu->page_table) << IO_PAGE_SHIFT); + + dma_base = iommu->page_table_map_base + + ((base - iommu->page_table) << IO_PAGE_SHIFT); /* Step 3: Normalize DMA addresses. */ used = nelems; @@ -702,30 +610,28 @@ int pci_map_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, int } used = nelems - used; - /* Step 4: Choose a context if necessary. */ - ctx = 0; - if (iommu->iommu_ctxflush) - ctx = iommu_alloc_ctx(iommu); - - /* Step 5: Create the mappings. */ + /* Step 4: Create the mappings. */ if (strbuf->strbuf_enabled) iopte_protection = IOPTE_STREAMING(ctx); else iopte_protection = IOPTE_CONSISTENT(ctx); if (direction != PCI_DMA_TODEVICE) iopte_protection |= IOPTE_WRITE; - fill_sg (base, sglist, used, nelems, iopte_protection); + + fill_sg(base, sglist, used, nelems, iopte_protection); + #ifdef VERIFY_SG verify_sglist(sglist, nelems, base, npages); #endif - spin_unlock_irqrestore(&iommu->lock, flags); - return used; bad: - spin_unlock_irqrestore(&iommu->lock, flags); - return PCI_DMA_ERROR_CODE; + iommu_free_ctx(iommu, ctx); +bad_no_ctx: + if (printk_ratelimit()) + WARN_ON(1); + return 0; } /* Unmap a set of streaming mode DMA translations. */ @@ -738,8 +644,10 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, unsigned long flags, ctx, i, npages; u32 bus_addr; - if (direction == PCI_DMA_NONE) - BUG(); + if (unlikely(direction == PCI_DMA_NONE)) { + if (printk_ratelimit()) + WARN_ON(1); + } pcp = pdev->sysdata; iommu = pcp->pbm->iommu; @@ -751,7 +659,8 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, if (sglist[i].dma_length == 0) break; i--; - npages = (IO_PAGE_ALIGN(sglist[i].dma_address + sglist[i].dma_length) - bus_addr) >> IO_PAGE_SHIFT; + npages = (IO_PAGE_ALIGN(sglist[i].dma_address + sglist[i].dma_length) - + bus_addr) >> IO_PAGE_SHIFT; base = iommu->page_table + ((bus_addr - iommu->page_table_map_base) >> IO_PAGE_SHIFT); @@ -772,11 +681,11 @@ void pci_unmap_sg(struct pci_dev *pdev, struct scatterlist *sglist, int nelems, if (strbuf->strbuf_enabled) pci_strbuf_flush(strbuf, iommu, bus_addr, ctx, npages, direction); - /* Step 2: Clear out first TSB entry. */ - iopte_make_dummy(iommu, base); + /* Step 2: Clear out the TSB entries. */ + for (i = 0; i < npages; i++) + iopte_make_dummy(iommu, base + i); - free_streaming_cluster(iommu, bus_addr - iommu->page_table_map_base, - npages, ctx); + free_npages(iommu, bus_addr - iommu->page_table_map_base, npages); iommu_free_ctx(iommu, ctx); diff --git a/include/asm-sparc64/pbm.h b/include/asm-sparc64/pbm.h index c067407de0b0..dd35a2c7798a 100644 --- a/include/asm-sparc64/pbm.h +++ b/include/asm-sparc64/pbm.h @@ -27,23 +27,27 @@ * PCI bus. */ -#define PBM_LOGCLUSTERS 3 -#define PBM_NCLUSTERS (1 << PBM_LOGCLUSTERS) - struct pci_controller_info; /* This contains the software state necessary to drive a PCI * controller's IOMMU. */ +struct pci_iommu_arena { + unsigned long *map; + unsigned int hint; + unsigned int limit; +}; + struct pci_iommu { /* This protects the controller's IOMMU and all * streaming buffers underneath. */ spinlock_t lock; + struct pci_iommu_arena arena; + /* IOMMU page table, a linear array of ioptes. */ iopte_t *page_table; /* The page table itself. */ - int page_table_sz_bits; /* log2 of ow many pages does it map? */ /* Base PCI memory space address where IOMMU mappings * begin. @@ -62,12 +66,6 @@ struct pci_iommu { */ unsigned long write_complete_reg; - /* The lowest used consistent mapping entry. Since - * we allocate consistent maps out of cluster 0 this - * is relative to the beginning of closter 0. - */ - u32 lowest_consistent_map; - /* In order to deal with some buggy third-party PCI bridges that * do wrong prefetching, we never mark valid mappings as invalid. * Instead we point them at this dummy page. @@ -75,16 +73,6 @@ struct pci_iommu { unsigned long dummy_page; unsigned long dummy_page_pa; - /* If PBM_NCLUSTERS is ever decreased to 4 or lower, - * or if largest supported page_table_sz * 8K goes above - * 2GB, you must increase the size of the type of - * these counters. You have been duly warned. -DaveM - */ - struct { - u16 next; - u16 flush; - } alloc_info[PBM_NCLUSTERS]; - /* CTX allocation. */ unsigned long ctx_lowest_free; unsigned long ctx_bitmap[IOMMU_NUM_CTXS / (sizeof(unsigned long) * 8)]; -- cgit v1.2.3-71-gd317 From 6205d158d16d2619bf30f0aff47a8e09b07106e9 Mon Sep 17 00:00:00 2001 From: Ben Dooks Date: Fri, 14 Oct 2005 12:24:24 +0100 Subject: [ARM] 3009/1: S3C2410 - io.h offsets too large for LDRH/STRH Patch from Ben Dooks The __inwc/__outwc calls are capable of creating LDRH and STRH instructions with offsets over 8bits as GCC does not have a constraint for an 8bit offset. Signed-off-by: Ben Dooks Signed-off-by: Russell King --- include/asm-arm/arch-s3c2410/io.h | 58 ++++++++++++++++++++++++++++----------- 1 file changed, 42 insertions(+), 16 deletions(-) (limited to 'include') diff --git a/include/asm-arm/arch-s3c2410/io.h b/include/asm-arm/arch-s3c2410/io.h index 418233a7ee6f..4bf272ed9add 100644 --- a/include/asm-arm/arch-s3c2410/io.h +++ b/include/asm-arm/arch-s3c2410/io.h @@ -9,7 +9,7 @@ * 06-Dec-1997 RMK Created. * 02-Sep-2003 BJD Modified for S3C2410 * 10-Mar-2005 LCVR Changed S3C2410_VA to S3C24XX_VA - * + * 13-Oct-2005 BJD Fixed problems with LDRH/STRH offset range */ #ifndef __ASM_ARM_ARCH_IO_H @@ -97,7 +97,7 @@ DECLARE_IO(int,l,"") else \ __asm__ __volatile__( \ "strb %0, [%1, #0] @ outbc" \ - : : "r" (value), "r" ((port))); \ + : : "r" (value), "r" ((port))); \ }) #define __inbc(port) \ @@ -110,35 +110,61 @@ DECLARE_IO(int,l,"") else \ __asm__ __volatile__( \ "ldrb %0, [%1, #0] @ inbc" \ - : "=r" (result) : "r" ((port))); \ + : "=r" (result) : "r" ((port))); \ result; \ }) #define __outwc(value,port) \ ({ \ unsigned long v = value; \ - if (__PORT_PCIO((port))) \ - __asm__ __volatile__( \ - "strh %0, [%1, %2] @ outwc" \ - : : "r" (v), "r" (PCIO_BASE), "Jr" ((port))); \ - else \ + if (__PORT_PCIO((port))) { \ + if ((port) < 256 && (port) > -256) \ + __asm__ __volatile__( \ + "strh %0, [%1, %2] @ outwc" \ + : : "r" (v), "r" (PCIO_BASE), "Jr" ((port))); \ + else if ((port) > 0) \ + __asm__ __volatile__( \ + "strh %0, [%1, %2] @ outwc" \ + : : "r" (v), \ + "r" (PCIO_BASE + ((port) & ~0xff)), \ + "Jr" (((port) & 0xff))); \ + else \ + __asm__ __volatile__( \ + "strh %0, [%1, #0] @ outwc" \ + : : "r" (v), \ + "r" (PCIO_BASE + (port))); \ + } else \ __asm__ __volatile__( \ "strh %0, [%1, #0] @ outwc" \ - : : "r" (v), "r" ((port))); \ + : : "r" (v), "r" ((port))); \ }) #define __inwc(port) \ ({ \ unsigned short result; \ - if (__PORT_PCIO((port))) \ - __asm__ __volatile__( \ - "ldrh %0, [%1, %2] @ inwc" \ - : "=r" (result) : "r" (PCIO_BASE), "Jr" ((port))); \ - else \ + if (__PORT_PCIO((port))) { \ + if ((port) < 256 && (port) > -256 ) \ + __asm__ __volatile__( \ + "ldrh %0, [%1, %2] @ inwc" \ + : "=r" (result) \ + : "r" (PCIO_BASE), \ + "Jr" ((port))); \ + else if ((port) > 0) \ + __asm__ __volatile__( \ + "ldrh %0, [%1, %2] @ inwc" \ + : "=r" (result) \ + : "r" (PCIO_BASE + ((port) & ~0xff)), \ + "Jr" (((port) & 0xff))); \ + else \ + __asm__ __volatile__( \ + "ldrh %0, [%1, #0] @ inwc" \ + : "=r" (result) \ + : "r" (PCIO_BASE + ((port)))); \ + } else \ __asm__ __volatile__( \ "ldrh %0, [%1, #0] @ inwc" \ - : "=r" (result) : "r" ((port))); \ - result; \ + : "=r" (result) : "r" ((port))); \ + result; \ }) #define __outlc(value,port) \ -- cgit v1.2.3-71-gd317 From cb38c569e5ecf9e922e66963b6da2751b4f13d81 Mon Sep 17 00:00:00 2001 From: Richard Purdie Date: Fri, 14 Oct 2005 16:07:25 +0100 Subject: [ARM] 3011/1: pxafb: Add ability to set device parent + fix spitz compile error Patch from Richard Purdie Add a function to allow machines to set the parent of the pxa framebuffer device. This means the power up/down sequence can be controlled where required by the machine. Update spitz to use the new function, fixing a compile error. Signed-off-by: Richard Purdie Signed-off-by: Russell King --- arch/arm/mach-pxa/generic.c | 5 +++++ arch/arm/mach-pxa/spitz.c | 2 +- include/asm-arm/arch-pxa/pxafb.h | 1 + 3 files changed, 7 insertions(+), 1 deletion(-) (limited to 'include') diff --git a/arch/arm/mach-pxa/generic.c b/arch/arm/mach-pxa/generic.c index d0660a8c4b70..d327c127eddb 100644 --- a/arch/arm/mach-pxa/generic.c +++ b/arch/arm/mach-pxa/generic.c @@ -208,6 +208,11 @@ static struct platform_device pxafb_device = { .resource = pxafb_resources, }; +void __init set_pxa_fb_parent(struct device *parent_dev) +{ + pxafb_device.dev.parent = parent_dev; +} + static struct platform_device ffuart_device = { .name = "pxa2xx-uart", .id = 0, diff --git a/arch/arm/mach-pxa/spitz.c b/arch/arm/mach-pxa/spitz.c index 568afe3d6e1a..7eaeb24aae1e 100644 --- a/arch/arm/mach-pxa/spitz.c +++ b/arch/arm/mach-pxa/spitz.c @@ -328,7 +328,7 @@ static void __init common_init(void) platform_add_devices(devices, ARRAY_SIZE(devices)); pxa_set_mci_info(&spitz_mci_platform_data); - pxafb_device.dev.parent = &spitzssp_device.dev; + set_pxa_fb_parent(&spitzssp_device.dev); set_pxa_fb_info(&spitz_pxafb_info); } diff --git a/include/asm-arm/arch-pxa/pxafb.h b/include/asm-arm/arch-pxa/pxafb.h index 21c0e16dce5f..aba9b30f4249 100644 --- a/include/asm-arm/arch-pxa/pxafb.h +++ b/include/asm-arm/arch-pxa/pxafb.h @@ -66,4 +66,5 @@ struct pxafb_mach_info { }; void set_pxa_fb_info(struct pxafb_mach_info *hard_pxa_fb_info); +void set_pxa_fb_parent(struct device *parent_dev); unsigned long pxafb_get_hsync_time(struct device *dev); -- cgit v1.2.3-71-gd317 From e26148d934762b61133a64b6862f870624ff617d Mon Sep 17 00:00:00 2001 From: Tim Schmielau Date: Fri, 14 Oct 2005 15:59:05 -0700 Subject: [PATCH] Fix copy-and-paste error in BSD accounting Fix copy and paste error in jiffies_to_AHZ conversion which leads to wrong BSD accounting information on alpha and ia64 when CONFIG_BSD_PROCESS_ACCT_V3 is turned on. Also update comment to match reorganised header files. Signed-off-by: Tim Schmielau Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/acct.h | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/linux/acct.h b/include/linux/acct.h index 1993a3691768..19f70462b3be 100644 --- a/include/linux/acct.h +++ b/include/linux/acct.h @@ -162,13 +162,13 @@ typedef struct acct acct_t; #ifdef __KERNEL__ /* * Yet another set of HZ to *HZ helper functions. - * See for the original. + * See for the original. */ static inline u32 jiffies_to_AHZ(unsigned long x) { #if (TICK_NSEC % (NSEC_PER_SEC / AHZ)) == 0 - return x / (HZ / USER_HZ); + return x / (HZ / AHZ); #else u64 tmp = (u64)x * TICK_NSEC; do_div(tmp, (NSEC_PER_SEC / AHZ)); -- cgit v1.2.3-71-gd317 From 688ce17b8599abc548b406c00e4d18ae0dec954f Mon Sep 17 00:00:00 2001 From: Al Viro Date: Sun, 16 Oct 2005 00:17:33 -0700 Subject: [PATCH]: highest_possible_processor_id() has to be a macro ... otherwise, things like alpha and sparc64 break and break badly. They define cpu_possible_map to something else in smp.h *AFTER* having included cpumask.h. If that puppy is a macro, expansion will happen at the actual caller, when we'd already seen #define cpu_possible_map ... and we will get the right thing used. As an inline helper it will be tokenized before we get to that define and that's it; no matter what we define later, it won't affect anything. We get modules with dependency on cpu_possible_map instead of the right symbol (phys_cpu_present_map in case of sparc64), or outright link errors if they are built-in. Signed-off-by: Al Viro Signed-off-by: David S. Miller --- include/linux/cpumask.h | 16 +++++++--------- 1 file changed, 7 insertions(+), 9 deletions(-) (limited to 'include') diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h index fe9778301d07..9bdba8169b41 100644 --- a/include/linux/cpumask.h +++ b/include/linux/cpumask.h @@ -393,15 +393,13 @@ extern cpumask_t cpu_present_map; #define for_each_present_cpu(cpu) for_each_cpu_mask((cpu), cpu_present_map) /* Find the highest possible smp_processor_id() */ -static inline unsigned int highest_possible_processor_id(void) -{ - unsigned int cpu, highest = 0; - - for_each_cpu_mask(cpu, cpu_possible_map) - highest = cpu; - - return highest; -} +#define highest_possible_processor_id() \ +({ \ + unsigned int cpu, highest = 0; \ + for_each_cpu_mask(cpu, cpu_possible_map) \ + highest = cpu; \ + highest; \ +}) #endif /* __LINUX_CPUMASK_H */ -- cgit v1.2.3-71-gd317 From b24d18aa743dad0c42918157c5d717686269d3a8 Mon Sep 17 00:00:00 2001 From: Herbert Xu Date: Sun, 16 Oct 2005 20:29:20 -0700 Subject: [PATCH] list: add missing rcu_dereference on first element It seems that all the list_*_rcu primitives are missing a memory barrier on the very first dereference. For example, #define list_for_each_rcu(pos, head) \ for (pos = (head)->next; prefetch(pos->next), pos != (head); \ pos = rcu_dereference(pos->next)) It will go something like: pos = (head)->next prefetch(pos->next) pos != (head) do stuff We're missing a barrier here. pos = rcu_dereference(pos->next) fetch pos->next barrier given by rcu_dereference(pos->next) store pos Without the missing barrier, the pos->next value may turn out to be stale. In fact, if "do stuff" were also dereferencing pos and relying on list_for_each_rcu to provide the barrier then it may also break. So here is a patch to make sure that we have a barrier for the first element in the list. Signed-off-by: Herbert Xu Acked-by: "Paul E. McKenney" Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- include/linux/list.h | 39 ++++++++++++++++++++++----------------- 1 file changed, 22 insertions(+), 17 deletions(-) (limited to 'include') diff --git a/include/linux/list.h b/include/linux/list.h index e6ec59682274..084971f333fe 100644 --- a/include/linux/list.h +++ b/include/linux/list.h @@ -442,12 +442,14 @@ static inline void list_splice_init(struct list_head *list, * as long as the traversal is guarded by rcu_read_lock(). */ #define list_for_each_rcu(pos, head) \ - for (pos = (head)->next; prefetch(pos->next), pos != (head); \ - pos = rcu_dereference(pos->next)) + for (pos = (head)->next; \ + prefetch(rcu_dereference(pos)->next), pos != (head); \ + pos = pos->next) #define __list_for_each_rcu(pos, head) \ - for (pos = (head)->next; pos != (head); \ - pos = rcu_dereference(pos->next)) + for (pos = (head)->next; \ + rcu_dereference(pos) != (head); \ + pos = pos->next) /** * list_for_each_safe_rcu - iterate over an rcu-protected list safe @@ -461,8 +463,9 @@ static inline void list_splice_init(struct list_head *list, * as long as the traversal is guarded by rcu_read_lock(). */ #define list_for_each_safe_rcu(pos, n, head) \ - for (pos = (head)->next, n = pos->next; pos != (head); \ - pos = rcu_dereference(n), n = pos->next) + for (pos = (head)->next; \ + n = rcu_dereference(pos)->next, pos != (head); \ + pos = n) /** * list_for_each_entry_rcu - iterate over rcu list of given type @@ -474,11 +477,11 @@ static inline void list_splice_init(struct list_head *list, * the _rcu list-mutation primitives such as list_add_rcu() * as long as the traversal is guarded by rcu_read_lock(). */ -#define list_for_each_entry_rcu(pos, head, member) \ - for (pos = list_entry((head)->next, typeof(*pos), member); \ - prefetch(pos->member.next), &pos->member != (head); \ - pos = rcu_dereference(list_entry(pos->member.next, \ - typeof(*pos), member))) +#define list_for_each_entry_rcu(pos, head, member) \ + for (pos = list_entry((head)->next, typeof(*pos), member); \ + prefetch(rcu_dereference(pos)->member.next), \ + &pos->member != (head); \ + pos = list_entry(pos->member.next, typeof(*pos), member)) /** @@ -492,8 +495,9 @@ static inline void list_splice_init(struct list_head *list, * as long as the traversal is guarded by rcu_read_lock(). */ #define list_for_each_continue_rcu(pos, head) \ - for ((pos) = (pos)->next; prefetch((pos)->next), (pos) != (head); \ - (pos) = rcu_dereference((pos)->next)) + for ((pos) = (pos)->next; \ + prefetch(rcu_dereference((pos))->next), (pos) != (head); \ + (pos) = (pos)->next) /* * Double linked lists with a single pointer list head. @@ -696,8 +700,9 @@ static inline void hlist_add_after_rcu(struct hlist_node *prev, pos = n) #define hlist_for_each_rcu(pos, head) \ - for ((pos) = (head)->first; pos && ({ prefetch((pos)->next); 1; }); \ - (pos) = rcu_dereference((pos)->next)) + for ((pos) = (head)->first; \ + rcu_dereference((pos)) && ({ prefetch((pos)->next); 1; }); \ + (pos) = (pos)->next) /** * hlist_for_each_entry - iterate over list of given type @@ -762,9 +767,9 @@ static inline void hlist_add_after_rcu(struct hlist_node *prev, */ #define hlist_for_each_entry_rcu(tpos, pos, head, member) \ for (pos = (head)->first; \ - pos && ({ prefetch(pos->next); 1;}) && \ + rcu_dereference(pos) && ({ prefetch(pos->next); 1;}) && \ ({ tpos = hlist_entry(pos, typeof(*tpos), member); 1;}); \ - pos = rcu_dereference(pos->next)) + pos = pos->next) #else #warning "don't include kernel headers in userspace" -- cgit v1.2.3-71-gd317 From 5ee832dbc6770135ec8d63296af0a4374557bb79 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Mon, 17 Oct 2005 20:01:21 +0200 Subject: [PATCH] rcu: keep rcu callback event counter This makes call_rcu() keep track of how many events there are on the RCU list, and cause a reschedule event when the list gets too long. This helps keep RCU event lists down. Signed-off-by: Linus Torvalds --- include/linux/rcupdate.h | 1 + kernel/rcupdate.c | 11 +++++++++++ 2 files changed, 12 insertions(+) (limited to 'include') diff --git a/include/linux/rcupdate.h b/include/linux/rcupdate.h index 4e65eb44adfd..70191a5a148f 100644 --- a/include/linux/rcupdate.h +++ b/include/linux/rcupdate.h @@ -94,6 +94,7 @@ struct rcu_data { long batch; /* Batch # for current RCU batch */ struct rcu_head *nxtlist; struct rcu_head **nxttail; + long count; /* # of queued items */ struct rcu_head *curlist; struct rcu_head **curtail; struct rcu_head *donelist; diff --git a/kernel/rcupdate.c b/kernel/rcupdate.c index dd99415b1551..2559d4b8f23f 100644 --- a/kernel/rcupdate.c +++ b/kernel/rcupdate.c @@ -109,6 +109,10 @@ void fastcall call_rcu(struct rcu_head *head, rdp = &__get_cpu_var(rcu_data); *rdp->nxttail = head; rdp->nxttail = &head->next; + + if (unlikely(++rdp->count > 10000)) + set_need_resched(); + local_irq_restore(flags); } @@ -140,6 +144,12 @@ void fastcall call_rcu_bh(struct rcu_head *head, rdp = &__get_cpu_var(rcu_bh_data); *rdp->nxttail = head; rdp->nxttail = &head->next; + rdp->count++; +/* + * Should we directly call rcu_do_batch() here ? + * if (unlikely(rdp->count > 10000)) + * rcu_do_batch(rdp); + */ local_irq_restore(flags); } @@ -157,6 +167,7 @@ static void rcu_do_batch(struct rcu_data *rdp) next = rdp->donelist = list->next; list->func(list); list = next; + rdp->count--; if (++count >= maxbatch) break; } -- cgit v1.2.3-71-gd317 From 4faa5285283fad081443e3612ca426a311bb6c7e Mon Sep 17 00:00:00 2001 From: Zach Brown Date: Mon, 17 Oct 2005 16:43:33 -0700 Subject: [PATCH] aio: revert lock_kiocb() lock_kiocb() was introduced to serialize retrying and cancellation. In the process of doing so it tried to sleep waiting for KIF_LOCKED while holding the ctx_lock spinlock. Recent fixes have ensured that multiple concurrent retries won't be attempted for a given iocb. Cancel has other problems and has no significant in-tree users that have been complaining about it. So for the immediate future we'll revert sleeping with the lock held and will address proper cancellation and retry serialization in the future. Signed-off-by: Zach Brown Acked-by: Benjamin LaHaise Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/aio.c | 26 +------------------------- include/linux/aio.h | 7 ++++++- 2 files changed, 7 insertions(+), 26 deletions(-) (limited to 'include') diff --git a/fs/aio.c b/fs/aio.c index d6b1551342b7..9fe7216457d8 100644 --- a/fs/aio.c +++ b/fs/aio.c @@ -398,7 +398,7 @@ static struct kiocb fastcall *__aio_get_req(struct kioctx *ctx) if (unlikely(!req)) return NULL; - req->ki_flags = 1 << KIF_LOCKED; + req->ki_flags = 0; req->ki_users = 2; req->ki_key = 0; req->ki_ctx = ctx; @@ -547,25 +547,6 @@ struct kioctx *lookup_ioctx(unsigned long ctx_id) return ioctx; } -static int lock_kiocb_action(void *param) -{ - schedule(); - return 0; -} - -static inline void lock_kiocb(struct kiocb *iocb) -{ - wait_on_bit_lock(&iocb->ki_flags, KIF_LOCKED, lock_kiocb_action, - TASK_UNINTERRUPTIBLE); -} - -static inline void unlock_kiocb(struct kiocb *iocb) -{ - kiocbClearLocked(iocb); - smp_mb__after_clear_bit(); - wake_up_bit(&iocb->ki_flags, KIF_LOCKED); -} - /* * use_mm * Makes the calling kernel thread take on the specified @@ -796,9 +777,7 @@ static int __aio_run_iocbs(struct kioctx *ctx) * Hold an extra reference while retrying i/o. */ iocb->ki_users++; /* grab extra reference */ - lock_kiocb(iocb); aio_run_iocb(iocb); - unlock_kiocb(iocb); if (__aio_put_req(ctx, iocb)) /* drop extra ref */ put_ioctx(ctx); } @@ -1542,7 +1521,6 @@ int fastcall io_submit_one(struct kioctx *ctx, struct iocb __user *user_iocb, spin_lock_irq(&ctx->ctx_lock); aio_run_iocb(req); - unlock_kiocb(req); if (!list_empty(&ctx->run_list)) { /* drain the run list */ while (__aio_run_iocbs(ctx)) @@ -1674,7 +1652,6 @@ asmlinkage long sys_io_cancel(aio_context_t ctx_id, struct iocb __user *iocb, if (NULL != cancel) { struct io_event tmp; pr_debug("calling cancel\n"); - lock_kiocb(kiocb); memset(&tmp, 0, sizeof(tmp)); tmp.obj = (u64)(unsigned long)kiocb->ki_obj.user; tmp.data = kiocb->ki_user_data; @@ -1686,7 +1663,6 @@ asmlinkage long sys_io_cancel(aio_context_t ctx_id, struct iocb __user *iocb, if (copy_to_user(result, &tmp, sizeof(tmp))) ret = -EFAULT; } - unlock_kiocb(kiocb); } else ret = -EINVAL; diff --git a/include/linux/aio.h b/include/linux/aio.h index 60def658b246..0decf66117c1 100644 --- a/include/linux/aio.h +++ b/include/linux/aio.h @@ -24,7 +24,12 @@ struct kioctx; #define KIOCB_SYNC_KEY (~0U) /* ki_flags bits */ -#define KIF_LOCKED 0 +/* + * This may be used for cancel/retry serialization in the future, but + * for now it's unused and we probably don't want modules to even + * think they can use it. + */ +/* #define KIF_LOCKED 0 */ #define KIF_KICKED 1 #define KIF_CANCELLED 2 -- cgit v1.2.3-71-gd317